[openstack-dev] [nova] When can/should we change additionalProperties=False in GET /servers(/detail)?
mriedemos at gmail.com
Wed Sep 19 14:41:45 UTC 2018
On 9/18/2018 12:26 PM, Matt Riedemann wrote:
> On 9/17/2018 9:41 PM, Ghanshyam Mann wrote:
>> ---- On Tue, 18 Sep 2018 09:33:30 +0900 Alex Xu<soulxu at gmail.com>
>> wrote ----
>> > That only means after 599276 we only have servers API and
>> os-instance-action API stopped accepting the undefined query parameter.
>> > What I'm thinking about is checking all the APIs, add
>> json-query-param checking with additionalProperties=True if the API
>> don't have yet. And using another microversion set
>> additionalProperties to False, then the whole Nova API become consistent.
>> I too vote for doing it for all other API together. Restricting the
>> unknown query or request param are very useful for API consistency.
>> Item#1 in this etherpadhttps://etherpad.openstack.org/p/nova-api-cleanup
>> If you would like, i can propose a quick spec for that and positive
>> response to do all together then we skip to do that in 599276
>> otherwise do it for GET servers in 599276.
> I don't care too much about changing all of the other
> additionalProperties=False in a single microversion given we're already
> kind of inconsistent with this in a few APIs. Consistency is ideal, but
> I thought we'd be lumping in other cleanups from the etherpad into the
> same microversion/spec which will likely slow it down during spec
> review. For example, I'd really like to get rid of the weird server
> response field prefixes like "OS-EXT-SRV-ATTR:". Would we put those into
> the same mass cleanup microversion / spec or split them into individual
> microversions? I'd prefer not to see an explosion of microversions for
> cleaning up oddities in the API, but I could see how doing them all in a
> single microversion could be complicated.
Just an update on https://review.openstack.org/#/c/599276/ - the change
is approved. We left additionalProperties=True in the GET
/servers(/detail) APIs for consistency with 2.5 and 2.26, and for
expediency in just getting the otherwise pretty simple change approved.
More information about the OpenStack-dev